THE RELATIO SHIP BETWEE EAR I GS MA AGEME T A D
E TERPRISE RISK MA AGEME T: THE EFFECT OF ERM COMMITTEE
ABSTRACT
This paper provides initial evidence on the relation between Enterprise Risk Management (ERM) and earnings management (EM). Using the establishment of an ERM management committee that integrates into a firm’s management structure and complements a traditional board risk committee as our proxy, we first provide some statistically significant evidence that earnings management and earnings volatility are among the factors that determine whether a firm adopts ERM as a holistic risk management framework. We then document how the existence of an ERM committee affects accrual based (AEM) and transaction-based (TEM) earnings management and earnings properties in subsequent periods. We find that an ERM committee correlates negatively with aggressiveness, smoothing and earnings volatility, but positively with firm performance as measured by earnings level. Lastly, we provide additional evidence that firms with relatively high earnings volatility benefit relatively more from the EM-mitigating effect of ERM.
Keywords: Transactions earnings management, Accrual earnings management, Enterprise Risk Management, Earnings volatility
1
1. I TRODUCTIO
This paper provides initial evidence on the relation between Enterprise Risk Management
(ERM) and earnings management (EM). We identify firms that have established ERM
committees and document that these firms display less earnings smoothing and earnings
aggressiveness in subsequent periods than do the firms in a matched sample that do not have
such committee. We also provide some evidence that these firms show earnings with a lower
volatility and a higher level in subsequent periods than the matched firms. These findings are
of interest to financial statement users including regulators and investors who try to assess the
accounting quality of firms. By newly documenting a link between ERM and earnings
management, our paper opens several avenues for future research.
Enterprise Risk Management (ERM) is a conceptual construct that comprises the
management systems of companies to holistically and strategically manage their exposure to
all kinds of value-relevant risks. Earnings management refers to the managerial practice “to
alter financial reports to either mislead some stakeholders about the underlying economic performance … or to influence contractual outcomes” (Healy and Wahlen 1999) by means of
manipulating accounting accruals - accrual-based earnings management (AEM) – and real business activities - transaction-based manipulations (TEM). ERM and earnings management
each has individually attracted significant research interest in the accounting literature.
However, the possible interrelation between them has not yet been researched and is therefore
not yet well understood.
2 Our motivation to link ERM and earnings management is that they share some common practices and goals, although for different reasons and with different long-term effects on accounting earnings and firm value. For example, they both influence managers’ selection and optimization of business transactions and the magnitude and timing of accounting accruals. They both also target similar outcomes for accounting earnings, for example in form of a higher short-term level or a smoother stream over time. We intend to disentangle these perceived commonalities and the consequent interrelation between ERM and earnings management. Specifically, we ask whether establishing an ERM management committee reduces earnings management, both AEM and TEM, and results in lower earnings volatility and a higher earnings level.
The remaining sections of this paper review the literature (Section 2), develop the hypotheses
(Section 3), describe the sample selection and variable definition (Section 4), lay out the methodology (Section 5), discuss the results (Section 6), and present our conclusions and their implications for further research (Section 7).
2. E TERPRISE RISK MA AGEME T A D EAR I GS MA AGEME T
ERM has emerged as a popular corporate practice, with more and more firms preferring its
“holistic” approach over the traditional “silo-based” approach to risk management (e.g.,
Gates and Hexter 2005). Other reasons for its popularity are, for example, the increased demand for risk management by investors, regulators, stock markets, and rating agencies.
Some studies have also found that highly leveraged firms are more inclined to adopt ERM
3 than are lightly leveraged firms (Liebenberg and Hoyt 2003). Defined as “the discipline by which an organization assesses, controls, exploits, finances, and monitors risks from all sources” , ERM has the overarching goal of managing corporate risks firm-wide and with a strategic perspective “for the purpose of increasing the organization’s short- and long term value to its stakeholders” (Casualty Actuarial Society Committee on Enterprise Risk
Management 2003, p. 8, as quoted in Gordon, Loeb and Tseng (2009)). Arguments in favor
of ERM working toward this value objective have referred to, for example, a reduction in
expected costs related to tax payments, financial distress, underinvestment, asymmetric
information, and risk mitigation for non-diversified stakeholders (e.g. Meulbroek 2002).
Proponents of ERM claim that ERM is designed to enhance shareholder value; however, portfolio theory suggests that costly ERM implementation would be unwelcome by
shareholders who can use less costly diversification to eliminate idiosyncratic risk.
Accordingly, previous research has tested whether ERM actually achieves to increase firm
value 1(Hoyt & Liebenberg, 2011) and if so, by which means. While constrained by some difficulties to identify and develop an adequate empirical measure for ERM, studies have documented some support that ERM has multiple benefits for fundamental business activities and financial outcomes. For example, ERM has been found to assist managerial decision- making by making risks more explicit and transparent, foster cooperation and integration across different functions and divisions (Kleffner, Lee, & McGannon, 2003),serve to generate synergies between risk management activities (Miccolis and Shah, 2000; Cumming and
Hirtle, 2001; Lam, 2001; Meulbroek, 2002), safeguard the firm’s reputation as a driver of future performance (Fombrun, Gardberg, & Barnett, 2000), achieve strategic, operational, reporting, and compliance objectives (Gordon, Loeb, & Tseng, 2009),increase firm value
1There is a large academic literature that investigates how firm value depends on total risk. For a review of that literature, see René Stulz, Risk Management and Derivatives , Southwestern Publishing, 2002.
4 (McShane, Nair, & Rustambekov, 2011), translate into excess stock market returns (Gordon,
Loeb, & Tseng, 2009), reduce stock price volatility, lower the cost-of-capital, or increase capital efficiency (Beasley et al., 2008, Hoyt and Liebenberg, 2011, Miccolis and Shah, 2000;
Cumming and Hirtle, 2001; Lam, 2003; Meulbroek, 2002). The tenor of these findings complements that of other studies providing evidence that individual practices to manage risk associate positively with firm value, for example hedging with derivatives (Bartram, Brown,
& Conrad, 2009; Carter, Rogers & Simkins, 2006; Graham & Rogers, 2002; Nelson, Moffitt,
& Affleck-Graves, 2005). If such individual, “silo-based” risk management practices each already associate positively with firm value, then it appears reasonable to expect that integrating these practices and managing them “holistically” with a portfolio approach would associate positively with firm value, too.
Some few papers have provided initial evidence that ERM might positively associate with value-relevant accounting performance outcomes such as earnings levels and volatility, confirming Lam’s prediction that ERM would allow firms to “produce more consistent
business results” (Lam, 2003).2 According to these papers, ERM associates with higher return-on-assets (ROA) (Hoyt & Liebenberg, 2011 ); (Baxter, Bedard, Hoitash, & Yezegel,
2012) and lower earnings volatility (Liebenberg & Hoyt, 2003). However, another study
finds little evidence for changes in earnings level or volatility after the appointment of a
CRO, which is one possible empirical proxy for a firm initiating ERM (Pagach & Warr,
2010).
This last set of consequences of ERM for accounting earnings – level and volatility –
overlaps with some of the consequences of earnings management. Earnings management is
2The benefits of reducing earnings volatility include increasing managerial compensation and wealth, reducing corporate income tax, reducing the cost of debt, avoiding underinvestment and earnings surprises, and mitigating volatility caused by low diversification (Barton, 2001).
5 the managerial practice to opportunistically modify reported short-term accounting performance outcomes without positively affecting fundamental firm value. It occurs when managers use their judgment and discretion “in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting practices” (Healy and Wahlen, 1999).
Managers might refer to two different instruments to manage earnings. On the one hand, there is transaction-based earnings management (TEM), the manipulation of earnings through real business activities that achieves a reported economic performance that deviate from the performance that would be warranted by normal operational practices. Studies that directly examine TEM have concentrated mostly on price discounts, acceleration of sales, alterations in shipment schedules, scale-backs in research and development (R&D), and delays of maintenance (Healy and Wahlen, 1999; Fudenberg and Tirole, 1995), and Dechow and
Skinner (2000) Roychowdhury (2006). On the other hand, there is accounting (accrual-based) earnings management (AEM), the manipulation of earnings through the opportunistic usage of discretion in applying accounting policies. With AEM, managers do not intervene in real economic activities but make accrual decisions that they would otherwise not make was it not merely for achieving some desired accounting outcomes. Following Jones (1991), a large body of literature has developed and applied various empirical strategies and models to detect and measure AEM; the most popular models are summarized in Table A1 (annex). Common to these strategies is that they are constrained by the noisiness of available proxies and by the fact that earnings management is driven by unobservable intentions and done within accounting regulations.
6 The short-term accounting objectives of earnings management may be diverse. On the one hand, they may refer to meeting or beating certain benchmarks like earnings break-even
(Burgstahler and Dichev, 1997), analyst forecasts (Das and Zhang, 2003), or other short-term reporting goals (Chamberlain and Magliolo, 1995). On the other hand, they may refer to the properties of earnings like increasing their level (DeAngelo, 1988; Pourciau, 1993) or
reducing their volatility (Lambert, 1984; Demski, 1998; Kirschenheiter and Melumad, 2002);
following (Leuz, Nanda, & Wysocki, 2003), we will refer to these two latter phenomena as
earnings aggressiveness and earnings smoothness.
These two latter consequences of earnings management are congruent with the earnings
consequences of ERM, providing motivation to link the two to each other in this study. If
ERM and earnings management pursue similar objectives for accounting earnings, then they
might complement or substitute each other. Prior evidence on this possible interrelation is
somewhat scant, fragmented, inconclusive, and geared towards the board oversight over risk
and earnings management. Barton (2001) finds that managing financial risks, proxied by
derivatives’ notional amounts, partially substitutes for AEM. He also document that the
magnitude of discretionary accruals relates to risk-management-related benefits such as
increasing managerial compensation and wealth, reduced corporate income taxes and debt
financing costs, preempted underinvestment and earnings surprises, and mitigated volatility
caused by low diversification – all dimensions that are arguably also within the objective set
of ERM (Meulbroek 2002). Davidson et al. (2005) document for a sample of Australian firms
that some governance mechanisms primarily targeting compliance risks – a majority of non-
executive directors on the board and its audit committee – are significantly associated with a
lower likelihood of AEM while some control mechanisms – the voluntary establishment of an
internal audit function and the choice of auditor – are not. Others similarly provide evidence
7 that the composition of the board and its audit committee, i.e. two governance institutions tasked with risk oversight, deter earnings management by constraining the managerial propensity to engage in it (Xie, Davidson III, & DaDalt, 2003; Bedard & Johnstone, 2004;
Klein, 2006; Davidson, Goodwin-Stewart, & Kent, 2005). However, we are not aware of any study that directly investigates the interrelation between managerial activities of risk management and managerial activities of earnings management.
In our analyses, we take the constitution of an ERM management committee as our proxy of
ERM. This choice is based on our understanding that constituting such a committee is a stronger signal of a comprehensive, firm-wide ERM than are generically describing ERM initiatives (e.g. Gordon, Loeb and Tseng 2009; Hoyt and Liebenberg 2011) or appointing a
CRO (Liebenberg and Hoyt (2003). In a first step we find that leverage, size, risk and firm performance are the main determinants for companies to implement an ERM management
committee. In a second step, we find evidence of a generally negative effect of ERM on
earnings management and its instruments TEM and AEM in subsequent periods. A more
rigorous “matching on observables” analysis confirms the negative relationship for earnings
aggressiveness and earnings smoothing but finds a positive relationship for AEM and no
relationship with TEM as measured with the Jones model. Lastly, we find that implementing
ERM increases AEM in firms with high earnings volatility. In such firms, managers might
have a particularly strong incentive to engage in earnings smoothing. However, as ERM prevents the use of TEM for achieving a smoother earnings stream, it introduces an incentive
to engage in more AEM to achieve this goal.
We make several contributions to the emerging literature on the benefits and costs of ERM.
First, we support previous evidence on the determinants for implementing ERM. Second, we
8 document how ERM might contribute to shareholder value by mitigating opportunistic earnings management. Third, we refine the predictions about the performance effects of ERM by documenting the simultaneous effects on earnings volatility and earnings levels. Previous research has addressed only one of these two effects at a time. This improved specification may provide on explanation why prior research has found only muted effects of ERM on performance. To our knowledge, none of these observations has previously been documented as integrative as in our study.
3. HYPOTHESIS DEVELOPME T
ERM is an institutionalized governance and control system that complements traditional governance and control structures and processes. First, by creating comprehensive risk accountabilities and reporting mechanisms, it increases both the information about and the scrutiny of the business by the corresponding institutions on board level, such as risk and audit committees (Lam, 2003), and on management level, such as the internal audit function.
It hence improves detecting, monitoring and mitigating earnings management as a factor of risk. 3 For example, while audit committees are traditionally monitoring quality and regulatory compliance of financial reporting, risk committees are tasked with expanding the scope of monitoring to non-compliance risks of strategy and operations. Second, an ERM management committee might be implemented to signal commitment to regulatory compliance or to impose an element of self-monitoring that might be more effective in curbing earnings management than outside monitoring. Outside monitoring is inherently constrained by
3 AEM entails the risks associated with its reversal and TEM entails the risks associated with entering sub- optimal actions, foregoing value-enhancing transactions, or mistiming transactions. Both forms of earnings management might result in loss of reputation, legal penalties, and negative stock price reactions when detected or corrected. Accounting restatements to reverse AEM results in both short- and long term adverse stock market consequences. For example, announcing accounting misstatements results in negative one-day stock market returns of -9-10% (Feroz et al. 1991 and Dechow, Hutton and Sloan 1996) and improved stock market performance in the three-year post- detection period happens only for firms that improve their corporate governance (Farber 2005).
9 unobservable managerial intentions and information asymmetries about the optimality of business transactions or reasonableness of accruals.
Thus, earnings management as a risk-generating accounting practice and earnings volatility as an accounting manifestation of risk might be considered as variables that ERM intends to address and that might hence also be among the determinants of the decision to implement
ERM. We deal with this potential endogeneity by analyzing the dynamic interaction between
EM and ERM. In a first stage, we analyze whether EM practices and earnings volatility are among the set of determinants for establishing an ERM committee. We use the results of this analysis in a second stage in which we use matching estimators to test the impact of implementing an ERM committee on subsequent EM practices, earnings distribution, and firm performance.
We are ex ante neutral on the possible relationship between earnings management or earnings volatility and the decision to establish an ERM committee. The reason is that either relationship could be supported by established theories. On the one hand, the quality and transparency on financial information is becoming a potential risk for companies, especially since financial scandals such Enron and WorldCom. Neoclassical theories can explain the incentives for companies to signal commitment to regulatory compliance and financial reporting quality. Accordingly, firms with already low earnings management or low earnings volatility would be motivated to establish ERM committees to reinforce their commitment signal.On the other hand, stakeholder theory would suggest the opposite. Firms with high levels of earnings management or high earnings volatility might want to mitigate risky practices and reduce volatility by implementing an additional and more effective mechanism of risk control.
10
H1a (stakeholder): High EM (both AEM and TEM) will increase the probability of implementing an ERM committee.
H1b (signaling): Low EM (both AEM and TEM) will increase the probability of implementing an ERM committee.
H2a (stakeholder): High earnings volatility (EVOL) will increase the probability of implementing an ERM committee;
H2b (signaling): Low earnings volatility (EVOL) will increase the probability of implementing an ERM committee.
In a second stage we look at the effects of implementing the ERM committee on subsequent
EM and earnings distribution. Financial theory suggests that firms have traditionally been concerned in reducing the costs associated with conflicts of interest between owners and managers and between shareholders and bondholders, expected bankruptcy costs, and also the costs of regulatory scrutiny. Pressure from a range of sources (NYSE, 2004; SOX 2002) has been interpreted as an increased emphasis on transparency and completeness of disclosures of trend and other qualitative information. Outside monitoring might be less effective in detecting analyzing the optimality of business transactions or reasonableness of accruals. The incentives and the actual level of earnings management are hence difficult to detect by outsiders. This is particularly the case for TEM that is arguably less subject to monitoring by directors, auditors, regulators and other outside stakeholders (Kim & Sohn,
2013). Therefore, by making the possible real economic consequences of earnings management transparent to directors and managers, ERM creates awareness of its legal, economic or reputational risks and hence might attack the propensity for earnings
11 management at its root. In sum, we argue that ERM limits managerial opportunities to engage in earnings management and hence propose the following hypothesis:
H3 : Firms that establish an ERM committee in a given period will engage in H3a : less AEM H3b : less TEM in subsequent periods than do firms without such a committee.
Much of the ERM literature argues that the earnings benefit of ERM is a lower earnings
volatility due to reduced cross-sectional risks (Liebenberg & Hoyt, 2003), but has not yet provided any clear evidence of this effect. However, lower earnings volatility is not per se the objective of ERM. The ultimate benefit of ERM is that it supports managers in optimizing the firm’s overall risk-performance relation and enhancing firm value by means of better overall management, more coordinated management and loss avoidance (Pagach & Warr, 2010) as well as by preventing risks from aggregating across different sources (Liebenberg & Hoyt,
2003). Such optimization can happen along two dimensions: managers can either reduce risks for a given (target) level of performance or improve performance for a given (target) level of risk, respectively. Reducing risks translates into a lower volatility of earnings while improving performance leads to a higher average level of earnings, both effects crowding out the incentive for managers to engage in earnings management directed at the same purposes.
To illustrate, consider what we might call raw earnings (the hypothetical construct of earnings before both ERM and EM), risk-managed earnings (the hypothetical construct of earnings after ERM but before EM), and earnings-managed earnings (the observed construct of reported earnings after both ERM and EM). Risk-managed earnings display a lower volatility (higher level) of earnings than raw earnings. Put differently, we argue that conditional earnings volatility – holding the earnings level constant – and conditional
12 earnings level – holding the earnings volatility constant – might be better measures of the effects of ERM than unconditional earnings volatility. Therefore, if a manager desires to opportunistically manipulate earnings to reduce their volatility (increase their level), she needs to engage in less EM if risk-managed earnings are the point-of-departure as compared to the raw earnings. Accordingly:
H4 : Firms that establish an ERM committee in a given period will display either H4a : a lower earnings volatility or H4b : a lower earnings mean in subsequent periods than do firms without such a committee.
However, we also expect that ERM, just like any economic activity, is inevitably subject to a
cost-benefit tradeoff. The costs associated with the described benefits possibly introduce
some incentives for more earnings management. We aptly frame this cost-benefit tradeoff in
terms of the risk-return relation. Reducing earnings volatility (less risk) is expected to be
accompanied with an (weakly) adverse effect on average earnings (less return) and increasing
earnings level (more return) is expected to be accompanied with an (weakly) adverse effect
on volatility (more risk), respectively. Because lower levels of earnings are typically
undesirable for managers, these costs introduce an incentive to inflate earnings, counteracting
the decreased incentive and opportunities described above. In the second case, the managerial
incentive to inflate earnings is less prevalent but the incentive for smoothing persists.
We expect that these undesirable incentives are more prevalent in firms that display extreme
magnitudes of volatility and earning levels, i.e. high volatility and low earnings means.
Therefore, our last hypotheses follow on the previous argument that firms might follow a
strategy of either reducing earnings volatility or increasing their earnings level to optimize
13 their risk-performance relation as the consequence of a more explicit risk management. For example, a firm with high earnings volatility might engage in a better business management with better timed transactions.
H5 : Firms that display high earnings volatility and that establish an ERM committee in a given period will engage in H5a : more AEM H5b : more TEM in subsequent periods than do firms without such a committee.
The opposite argument applies to firms that are in a low-earnings spectrum. Implementing
ERM and incurring the associated costs would further depress their earnings. This is undesirable and hence introduces an incentive to engage in more earnings management to counteract this effect. We hence hypothesize:
H6 : Firms that display low earnings mean and that establish an ERM committee in a given period will engage in H6a : more AEM H6b : more TEM in subsequent periods than do firms without such a committee.
4. RESEARCH DESIG A D DATA
4.1. Sample
The ERM committee sample used for this study was derived from a Lexis-Nexis word search for companies that indicated they in their SEC filings they have an ERC committee or similar. Following Hoyt and Liebenberg (2009), firms were initially identified as having an
ERM committee based on a search of the following key terms: Strategic risk Management
14 Committee, Enterprise Risk Management Committee, Compliance and Operational Risk
Committee, Enterprise Risk Teams, Internal Risk Control Group, Corporate Financial Risk
Management Committee, Corporate Risk Committee, Enterprise Risk Council, Risk
Management Committee, Risk Committee of Management, Operational Risk Management
Committee, Operational Risk Management Committee, or Risk Working Group. The sentences that contain the key words were read to get a better sense of whether or not the
ERM concept is actually being used. Appendix A provides three examples of disclosures concerning the implementation of ERM committee in firms. Based on the keywords searching process, 531 US firms were identified as having created an ERM committee between 1993 and 2010, a total of 8,113 firm-year observations. We exclude 3,569 firm-year observations of financial institutions (SIC codes 6000-6999) because their accounting regulation differ too much from the rest of companies. Lexis-Nexis database searches across the complete population sample of Compustat companies, therefore the remaining companies in Compustat database, do not implement an ERMC. Total firm-year observations of the sample, with the main descriptive statistics, are summarized in Table 2.
4.2. Measures for Enterprise Risk Management and Earnings Management
4.2.1. Enterprise Risk Management Measure
A major obstacle to empirical ERM-related research is the difficulty in identifying firms that
are indeed engaging in ERM. Researchers find difficult to detect if the firm is managing risks
in a disaggregated or aggregated manner. Thus they are forced to either rely on survey data or
search for a signal of the existence of ERM programs. One such signal may come from the
creation of a specialized managerial position, the Chief Risk Officer (CRO), which is
responsible for ERM implementation and coordination (Liebenberg & Hoyt, 2003). However
we understand that the existence of an Enterprise Risk Management Committee (ERMC), or
15 similar, is a better proxy for the implementation of a holistic risk management structure.
ERM is a dummy variable which takes value of 1 if the company has a risk committee at the managerial level and 0 otherwise.
4.2.2. Accrual-based Earnings Management measures
Prior studies on earnings management use different discretionary models to proxy for AEM.
Measures for AEM lack of power and are highly noisy by construction. We construct an
aggregate measure capturing earnings smoothing (Leuz et al, 2003), earnings aggressiveness
(Leuz et al, 2003) and discretionary accruals (Dechow, Sloan, & Sweeney, 1995).
• EM1: Earnings Smoothness (Leuz et al, 2003)
(σ(OPINC it/TA it-1)/ σ(CFO 4it/TA it-1) )
• EM3: Earnings Aggressiveness (Leuz et al, 2003)
[|ACCit |/ TA i,t-1] / [|CFOit| /TA i,t-1]5
• Jones: Modified Jones Model (Dechow et al, 1995)
ACC it / TAit-1= β0 + β1(1/TA it-1) + β2(∆SALE it /TA it-1–∆REC it /TA it-1) + β3PPE it /TA it-1+ εit ,
WhereOPINC is DATAXX from Compustat, CFO is DATAXX from Compustat, ACC is
defined as the change in non-cash current assets minus the change in current liabilities
excluding the current portion of long-term debt, minus depreciation and amortization, scaled by lagged total assets 6, CFO is DATAXX from Compustat, For Modified Jones Model ACCit
is computed as above, TAit-1 is lagged total assets, ∆SALE it is the change in sales, ∆REC it is the change in accounts receivable and PPE it is net property, plant, and equipment. We take
4The firm-level “rolling” standard deviations of operating income and operating cash flow both scaled by lagged total assets. 5Higher score simply more earnings mangement. 6 ACC = (∆total currentassets/TA – ∆cash/TA) - (∆total currentliabilities/TA - ∆short-termdebt/TA - ∆taxes payable/TA) - depreciationexpense/TA
16 the absolute value of the residual from the above regression as a measure of discretionary accruals of firm iat time t.
AEM1: Aggregate measure of Accrual Earnings Management
AEM1 is the mean of rank_em1re1, rank_em3 and rank_jones, where rank_em1re is the deviation of em1 from one, rank_em3 is the rank of em3 as in equation (x+2) and rank_jones is the rank of the residual absolute values of Jones as in equation (x+3). Rank_em1re1 is a proxy for earnings management considering both smoothness and aggressiveness.
AEM2 is the rank Jones measure
4.2.3. Transaction-based Earnings Management Measures
To proxy for Transaction earnings management (TEM) we follow the model developed by
Roychowdhury (2006), and implemented by Cohen, Dey and Lys (2008) that consider the abnormal levels of cash flow from operations (CFO), discretionary expenses and productions costs to proxy real activities manipulations. We focus on three manipulation methods and their impact on the above three variables: acceleration of the timing of sales through increased price discounts or more lenient credit terms; reporting of lower cost of goods sold through increased production, and decreases in discretionary expenses which include advertising expense, research and development, and SG&A expenses.
The normal levels of CFO, discretionary expenses and production costs are generated using the model developed by Dechow, Kothari and Watts (1998) as implemented in
Roychowdhury (2006). We express normal CFO as a linear function of sales and change in sales. To estimate this model, we run the following cross-sectional regression for each industry and year:
17 CFO it /TA it-1 = k1t (1/ASS it-1) + k 2(SALE it /ASS it-1) + k 3 (∆SALE it /TA it-1) + ε it
Abnormal CFO is actual CFO minus the normal level of CFO calculated using the estimated coefficient from (4).
Following Cohen, Dey and Lys (2008) production costs are defined as the sum of cost of goods sold (COGS) and change in inventory during the year.
COGS it /TA it-1= k 1t (1/TA it-1) + k 2SALE it /TA it-1 + ε it
∆INV it /TA it-1 = k 1t + (1/TA it-1) + k 2∆SALE it /TA it-1+ k 3∆SALE it-1/TA it-1 + ε it
Adding the previous equations the normal level of production costs is estimated as follows:
PROD it /TA it-1 = k 1t (1/TA it-1) + k 2SALE it /TA it-1+ k 3∆SALE it /TA it-1+ k 4∆SALE it-1/TA it-1 + ε it
The normal level of discretionary expenses is a linear function of current sales; however
Cohen et al (2008) address the fact that some companies might manage upwards reported earnings certain year, resulting in lower residuals by modeling discretionary expenses as a function of lagged sales 7:
DISCREX it /TA it = k 1t (1/TA it-1) + k 2SALE it-1 /TA it-1 + ε it
The abnormal levels of CFO, production costs and discretionary expenses are computed as
the difference between the values reported in COMPUSTAT and the levels predicted in previous equations. We rank these three variables in its absolute values and average them to
generate an aggregated measure as proxy for transaction earnings management.
7 CFO is cash flow from operations in period t (Compustat data item 308 – annual Compustat data item 124); Prod represents the production costs in period t, defined as the sum of COGS (annual Compustat data item 41) and the change in inventories (annual Compustat data item 3); DiscExp represents the discretionary expenditures in period t, defined as the sum of advertising expenses (annual Compustat data item 45), R&D expenses (annual Compustat data item 46)16 and SG&A (annual Compustat data item 189).
18 4.2.4. Control Variables
Beasley, Pagach and Warr, 2007 find that firms with less cash and more leverage are likely to see benefits from ERM. Furthermore, shareholders of large non-financial firms, with volatile earnings, low amounts of leverage and low amounts of cash on hand also react favorably to the implementation of ERM. These findings are consistent with the idea that a well implemented ERM program can create value when it reduces the likelihood of costly lower tail outcomes, such as financial distress.
Financial literature also relates ERM with firm value and use Tobin’s Q as a proxy for firm value since TQ compares the market value of a firm’s assets to their replacement cost. It has been used to measure (Yermack, 1996; Morck, Schleifer, and Vishny, 1988; (Hoyt &
Liebenberg, 2011). We will control for systematic risk using Beta KMV (BETA) (Acharya,
Almeida and Campello, (2012). This unlevered beta measures the amount of systematic risk
inherent in a firm's equity compared with the overall market.
We control for size to control for political costs related to size (SIZE), since it is often used to proxy for political costs (e.g. Watts & Zimmerman 1986). However, firm size could proxy
for factors other than political visibility such as information environment, capital market pressure, or financial resources. Several studies hypothesize fixed costs associated with
maintaining adequate internal control procedures, and hence predict a positive relation between firm size and internal control quality (Dechow, Ge, & Schrand, 2010). We control
for firm performance (ROA) since studies hypothesize that weak financial performance provides incentives for earnings management using the discretion allowed in accounting standard adoption to their advantage (Petroni, 1992; DeFond and Park, 1997; Balsam, Haw,
19 and Lilien, (1995 ); Keating and Zimmerman , 1999; Doyle, Ge, and McVay, 2007; Kinney and McDaniel, 1989). We control for growth (TQ) on the fundamental element of earnings properties; growth also is associated with greater measurement error and more manipulation opportunities (Richardson et al., 2005). We control for leverage (LEV) since several studies find evidence on positive relationship between highly levered firms and EM (Kinney and
McDaniel, 1989; Efendi et al., 2007; Dechow et al., 1996). Finally we control for GDP growth (GDP) to capture the effect of “good” times and “bad” times on EM. Table 1 provides detailed information on the measures for these control variables.
5. METHODOLOGY A D EMPIRICAL MODELS
5.1. Time-to-implementation models
Our first hypotheses postulate that the decision of implementing ERM is a function of
EVOL (H0a), EMEAN (H0b) and EM (H0c). In order to test the hypotheses, we set up a
“time-to-implement” discrete-time duration model, where we specify the hazard rate (the
probability of implementing ERM at time t conditional on not having implemented before
time t) is a logit function of time-varying variables. We estimate the following logit
model: